The idea is simple enough: we want to identify the patients with a resected early stage NSCLC that has a high risk of recurrence, so that we can give them additional therapy, usually in the form of post-operative (adjuvant) chemotherapy, while sparing this additional challenging and even potentially dangerous therapy for the people who have a more favorable prognosis.  Right now, the system we generally use to identify patients at higher vs. lower risk of recurrence is based on lung cancer staging, which is based on the size of the tumor, whether the cancer invades nearby structures, the extent of lymph node involvement, and whether there is spread to distant parts of the body.  That staging system works reasonably well, but we also know that there are many patients with early stage disease who fall below the general threshold for adjuvant chemotherapy but will still recur, while many people who receive chemotherapy were already cured from the surgery alone.  This leads us to the question, "Can we refine our prognostic ability by looking at the molecular characteristics of the cancer?".  We keep learning that molecular features of a cancer are becoming less opaque and are very important, and we also know that some tumors of the same stage respond much more aggressively than others.

We've looked at molecular markers (one or a few distinct variables that may be particularly important) and genetic signatures (an overall pattern of the genetic activity level of an often complex collection of genes) for a few years.  For instance, ERCC-1 is an enzyme that was shown in a retrospective analysis of a single trial was prognostic for survival and predictive of the benefit or lack of benefit from adjuvant cisplatin based chemo, and an analysis of a different adjuvant chemotherapy trial demonstrated that a molecular signature was also prognostic of better or worse survival and predictive of which patients benefit from chemotherapy.  But these methods have had limitations and detractors.  Markers that are called positive or negative, high or low by immunohistochemistry (IHC), which is essentially the intensity and frequency of a protein on the surface of cancer cells, is a subjective interpretation.  And work on molecular signatures has generally required "snap frozen" tissue, which requires special preparation right at the time of the surgery, and this is a technical challenge that is not widely feasible.

It is in this setting that the results of a study by Kratz and colleagues, now published online in the prestigious journal the Lancet, is novel and interesting.  This work was done at the University of California - San Francisco (UCSF) in collaboration with the company Pinpoint Genomics (conflict of interest: though I have no financial interest in these entities,  I have been friends for several years with both Dr. David Jablons, surgeon at UCSF who directed much of this work, and Mr. David Berryman, who is CEO of Pinpoint Genomics, though I knew him mostly from his time working at Genentech several years ago).  They set out to develop a "gene expression assay", a predictive pattern from multiple genes expressed by tumors, that could be run from formalin-fixed paraffin-embedded (FFPE, the basic way that tumor tissue is routinely stored after surgery everywhere), developing this from the pairing of gene expression patterns with the outcomes of 361 patients with stage I non-squamous NSCLC who had undergone surgery at UCSF and whose outcomes in the years after surgery were known.   They were able to develop a 14-gene assay that could separate patients into low, intermediate, or high risk groups (of approproximate 1/3 of the overall population each) with a 5 year overall survival (OS) of 71.4%, 58.3%, and 49.2%, respectively.

They then tested this multi-gene assay that was developed with the data from UCSF patients against two other groups.  One was pretty much next door, as they reviewed the prognostic utility of the assay on 433 patients with stage I non-squamous NSCLC from Kaiser Permanente Division of Research, centered in California; the second was a group of 1006 patients with stage I to III non-squamous NSCLC from a research group called the China Clinical Trials Consortium (CCTC).  For the Chinese patients, the same assay was done in a lab based in China, in order to test whether it would also work if the molecular assay is done in a different lab and on a cohort of patients with a very different genetic composition compared with the California-based populations.

The Pinpoint Genomics assay worked just as well at providing a helpful division of prognosis into low, intermediate, and high risk groups in the Kaiser (5-year OS: 71.4%, 58.3%, and 49.2%, respectively) and CCTC (5-year OS 74.1%, 57.4%, and 44.6%, respectively) cohorts.  They also compared how helpful the molecular assay is compared with other factors like age, sex, smoking status, histologic subtype (though squamous NSCLC was excluded, adenocarcinoma, large cell, mixed subtype, and NSCLC "not otherwise specified" were all included), and tumor size greater than 4 cm or not; this analysis revealed that being in the intermediate or especially high risk group was among the most predictive factors.  It was also more predictive than some a combination of factors that are advocated by the National Comprehensive Cancer Network (NCCN) guidelines as being of some utility in raising the concern for recurrence and consideration of chemotherapy in patients with smaller node-negative cancers: poorly differentiated tumor, vascular invasion, history of a wedge resection rather than lobectomy or bigger surgery, visceral pleural (the lining around the lung) involvement, and unknown lymph node status (not assessed properly).

Suffice it to say that this assay appears potentially valuable in helping guide us in our decision-making about who is at higher risk of recurrence and dying from their early stage cancer.  What are the limitations? First, this work is all retrospective.  We can identify who did poorly, but we don't know that we can identify patients prospectively and then enable them to do better because of an intervention.  We want to presume that if we can identify patients with stage I NSCLC who have a high risk profile, we can say that the anticipated benefit from adjuvant chemotherapy in reducing that risk of recurrence clearly outweighs the risk of serious problems from more treatment.  However, we don't know that: it could be that this test only identifies that patients will do less well but that we can't reduce that risk of recurrence by giving more treatment.  A trial is being conducted that is prospectively assigning stage I patients identified as high risk by this assay to either additional chemotherapy or observation, but it will literally be many years, and even likely a decade or more, before we have the results of this study (studies of early stage patients take many years to see whether the intervention made a difference 3-5 years later).  And this assay doesn't apply to the 20-25% of patients with a squamous NSCLC, who were presumably excluded because of the finding that the pattern of genetic derangements is too different from non-squamous NSCLC to pool these groups together and expect to see the same trends with the same gene collection.

In the meantime, the authors highlight several advances from this work.  This work represents a test that is more objective than an interpretation of IHC results, and they were able to show the assay's prognostic utility from different labs in different parts of the world, and looking at very large populations that were very different from one another.  In light of the growing work showing the differences in molecular features of tumors from Asian and North American or European populations, demonstrating that the same multi-gene assay produces very similar results in North American and Chinese patients is a meaningful accomplishment.

The authors note that the prospective research is underway but will take many years to complete and report.  In the meantime, they note that the NCCN advocates incorporating "soft" variables like tumor grade, pleural or blood vessel invasion, and adequacy of the surgery in the decision of whether to recommend adjuvant chemotherapy, and it's fair to highlight that this test is better studied and has more evidence behind it than the loose clinical considerations that the NCCN favors incorporating based on faith and presumptions alone.  However, saying that inadequate evidence is better than poor evidence is damning with faint praise.  (The NCCN guidelines are implied to be evidence-based, but that's only selective: they are a product of human effort that involves a combination of evidence and the biases of real people who still promote faith-based edicts.)

In the end, though, it's also fair to say that we are left facing decisions that often don't have perfect evidence to guide them.  The circumstances for individual patients and whether we would strongly recommend adjuvant chemotherapy are often fuzzy: many patients do have cancers that fall along the border of sufficient risk of recurrence to justify chemo, and other patients may have a cancer that merits chemotherapy by a pathologic staging-based system but may be riskier candidates for chemotherapy or just very wary about accepting it after a big lung surgery.  Having this additional information from a commercially available test could certainly serve as a useful additional piece of information for some patients and their doctors as they struggle to get the best sense of their probability of doing well and whether to accept the additional challenge of more treatment after resection.

I imagine that many/most patients would be very interested in having one more piece of prognostic information at their disposal. Would you favor having your tissue tested with this assay and getting a result back of low, intermediate, or high risk? Would you participate on a randomized trial that demonstrates that your stage I tumor is high risk but that then randomly assigns you to either chemotherapy or observation? I'm very interested in your thoughts.